Article Text
PDF
Abstract
Background This study comprehensively investigates the association between the expression of nicotinamide N-methyltransferase (NNMT) and clinical outcomes of urothelial bladder cancer (UBC), as well as the molecular mechanisms by which NNMT in cancer-associated fibroblast (CAF) modulates tumor progression and immunotherapy resistance in UBC.
Methods Single-cell transcriptomic analyses, immunohistochemical and immunofluorescence assays were performed on bladder cancer samples to validate the relationship between NNMT expression and clinical outcomes. A series of experiments, including chromatin immunoprecipitation assay, liquid chromatography tandem mass spectrometry assay, and CRISPR‒Cas9 (Clustered Regularly Interspaced Short Palindromic Repeats and CRISPR-associated protein 9) knockout, together with in vivo models, have been established to determine the molecular functions of NNMT in CAFs in UBC.
Results We demonstrated that elevated expression of the nicotinamide adenine dinucleotide (NAD+) metabolism enzyme NNMT in CAFs (NNMT+ CAFs) was significantly associated with non-response to programmed death-ligand 1 (PD-L1) blockade immunotherapy in patients with UBC and predicted the unfavorable prognosis of UBC in two independent large cohorts. Targeting NNMT using the inhibitor 5-Amino-1-methylquinolinium iodide significantly reduced tumor growth and enhanced the apoptotic effects of the anti-PD-L1 antibody in UBC mouse models. Mechanistically, NNMT+ CAFs recruit tumor-associated macrophages via epigenetic reprogramming of serum amyloid A (SAA) to drive tumor cell proliferation and confer resistance to programmed death-1/PD-L1 blockade immunotherapy.
Conclusions NNMT+ CAFs were significantly associated with non-response to PD-L1 blockade immunotherapy in patients with UBC. Elevated NNMT, specifically in CAFs, upregulates SAA expression and enhances the recruitment and differentiation of macrophages in the tumor microenvironment, thereby directly or indirectly promoting tumor progression and conferring resistance to immunotherapies in bladder cancer.
- Macrophage
- Tumor Microenvironment
- Bladder Cancer
- Immune Checkpoint Inhibitor
- Immunotherapy
Data availability statement
Data are available in a public, open access repository. Data sharing not applicable as no datasets generated and/or analyzed for this study.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Although robust and durable responses have been obtained by immune checkpoint inhibitors (ICIs) immunotherapy for urothelial bladder cancer (UBC), only a small subset of patients could benefit from ICI-based immunotherapy. Therefore, elucidating the determinants conferring resistance to immunotherapy is crucial and urgently needed.
WHAT THIS STUDY ADDS
Nicotinamide N-methyltransferase (NNMT+) cancer-associated fibroblasts (CAFs) are significantly associated with poor prognosis and non-response to programmed death-ligand 1 (PD-L1) blockade immunotherapy in patients with bladder cancer. By regulating macrophages infiltration and differentiation, the elevated expression of NNMT in CAFs promotes tumor progression and conferring resistance to immunotherapy in bladder cancer.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Our study revealed a novel mechanism of the NNMT+ CAFs-serum amyloid As-tumor-associated macrophages conferring resistance to programmed death-1/PD-L1 blockade immunotherapy in UCB. Targeting NNMT+ CAFs might provide a novel therapeutic approach for combinatory immunotherapies for bladder cancer using ICIs.
Introduction
Urothelial bladder cancer (UBC) is one of the most common malignant tumors of the urinary system in men.1 Progressive subtypes of UBC exhibit stromal cell and lymphocyte infiltration in the tumor microenvironment (TME). Although robust and durable responses have been obtained via immunotherapy for UBC, tumor recurrence is frequently observed with BCG treatment,2 and only a small subset of patients (15–25%) can benefit from immune checkpoint inhibitors (ICIs) such as programmed death-1/programmed dead-ligand 1 (PD-1/PD-L1) antibody immunotherapy.3 Therefore, elucidating the determinants of immunotherapy resistance is crucial as such studies will facilitate the development of new treatment strategies for bladder cancer.
The therapeutic challenges in immunotherapy can be attributed to complex interactions within the TME, in which multiple cellular compartments, including malignant cells and stromal, immune, and endothelial cells, provide plasticity for tumor survival.4 5 Cancer-associated fibroblasts (CAFs) are the main cell type constituting the stroma and contribute to tumor progression by modulating cancer metabolism.6 7 Emerging evidence suggests that altered metabolic pathways in cancer cells are key determinants of response to immunotherapy;8 9 however, the roles and mechanisms of metabolic reprogramming in cancer immunotherapy are poorly understood.
Previous studies have established that epigenetically-activated CAFs secrete numerous molecules and serve as organizations for interactions among different cells, thereby supporting the growth of malignant cells.6 CAFs may increase glucose consumption and lactate production to provide energy for tumor proliferation and invasion, which is considered the “reverse Warburg effect”.10 Moreover, inflammatory CAFs (iCAFs), distinct from other subgroups of CAFs such as myo-cancer-associated fibroblasts (mCAFs) and antigen-presenting cancer-associated fibroblasts (apCAFs), are required for tumor inflammation in the TME.11 Tumor inflammation can further induce cancer cell dedifferentiation and acquire resistance to cancer immunotherapy.12 Owing to the important roles of CAFs, multiple studies have shown that targeting CAFs is a promising approach for the treatment of cancers.13 14 A recent study on pancreatic ductal adenocarcinoma showed that LRRC15+ CAFs, a population of CAFs responsive to transforming growth factor-β (TGF-β) and expressing the leucine-rich repeat containing 15 (LRRC15) protein, were associated with a poor response to anti-PD-L1 immunotherapy.15 However, the diverse CAF populations in tumors and the uncharacterized mechanisms of fibroblast-mediated immunotherapy response present significant challenges in targeting CAFs for cancer treatment.
Tumor-associated macrophages (TAMs) have been documented to play a key role in bladder cancer progression.16 Pro-tumor macrophage infiltration in bladder tumors has been reported to be associated with poor prognosis and failure of BCG immunotherapy.17 In a previous study, we found that a subpopulation of TAMs (CD68, CD204, and CD206) was enriched in the tumor stroma of UBC and associated with unfavorable clinicopathological features.18 Owing to the dynamic co-evolution of TAMs and CAFs in tumors, elucidating the interactions between CAFs and TAMs may aid in understanding the mechanisms of tumor progression and the development of immunotherapy resistance in UBCs.
Epigenetic reprogramming of nicotinamide adenine dinucleotide (NAD+) metabolism has emerged as an important regulatory pathway in tumor progression and drug resistance.19 NAD+ synthesis connects substrate oxidation by the tricarboxylic acid cycle to energy production through the electron transport chain and oxidative phosphorylation, and regulates signaling pathways linked to cancers.20 Nicotinamide N-methyltransferase (NNMT) is a key NAD+ metabolism enzyme that catalyzes the N-methylation of nicotinamide and other pyridines using dose-dependent S-adenosyl methionine (SAM) as a methyl donor, which broadly alters methylation profiles by modifying the expression of cancer genes.21 Although previous studies have shown that NNMT is upregulated in tumors,22–24 the molecular mechanisms of NNMT in cancer development remain unknown. Elevated NNMT expression mediated by BRCA2 (breast cancer 2, early onset) depletion in ovarian cancer has been suggested to promote tumor progression.25 Additionally, NNMT in the tumor stroma could also exert its functions to mediate ovarian cancer migration, proliferation and metastasis.22 In addition to NNMT, a recent study by Lv et al suggested that in cancer cells, nicotinamide phosphoribosyl transferase, a rate-limiting enzyme in NAD+ biogenesis, contributes to tumor immune evasion by regulating interferon γ-Stat1 signaling to maintain PD-L1 expression.19 These studies highlight the importance of NAD+ metabolic enzymes in cancer development and immune regulation. However, only a few studies have investigated the clinical relevance of NNMT in UBC. Consequently, the roles and molecular mechanisms of NNMT in the immunology of UBC remain unknown.
Herein, we investigated the molecular mechanisms by which NNMT+ CAFs promote tumor progression and confer resistance to immunotherapy in patients with UBC. We comprehensively explored the biological functions of NNMT+ CAFs in UBC using clinical analyses of two large cohorts, in vitro experiments, and cancer mouse models, and elucidated the mechanism by which NNMT epigenetically reprogrammed CAFs regulate serum amyloid A (SAA) signaling in TAMs, thereby promoting tumor progression and conferring resistance to anti-PD-1 treatment. These findings may have a significant impact on the development of novel therapeutic approaches for UBC treatment.
Materials and methods
Human tissue specimens and patient clinical information
A total of four independent sets of UBC tissue samples were collected for studies from patients who underwent transurethral resection or cystectomy at Sun Yat-sen Memorial Hospital (Guangzhou, Guangdong, China) and Sun Yat-sen University Cancer Center (Guangzhou, Guangdong, China). The Sun Yat-sen Memorial Hospital (SYSMH) cohort (cohort A) included 177 formalin-fixed, paraffin-embedded (FFPE) tissue samples from patients who were hospitalized between July 2003 and July 2016. The Sun Yat-sen University Cancer Center (SYSUCC) cohort (cohort B) included 160 FFPE tissue samples from patients between February 2008 and November 2018. We used the same enrollment criteria for sample collection in both cohort A and cohort B. None of the patients had received any preoperative anticancer therapy. All tumors were independently graded by two pathologists according to the 2014 WHO classification system and staged according to the tumor-node-metastasis (TNM) classification system (eighth edition, 2016). The clinical information of the cohort A and cohort B were collected through a medical record review, and a follow-up survey was conducted to gather information on vital events via telephone interviews. The third independent cohort (cohort C) included 32 freshly frozen primary UBC specimens collected at the Sun Yat-Sen Memorial Hospital between June 2015 and January 2016, which were used to isolate total RNA to assess NNMT and SAA1 messenger RNA (mRNA) levels using quantitative reverse transcription PCR (RT-qPCR). The fourth independent cohort (cohort D) included 13 patients with UBC who were treated with four courses of neoadjuvant tislelizumab combined with cisplatin and gemcitabine at Sun Yat-sen Memorial Hospital between February and December 2021. We selected patients with advanced muscle-invasive bladder cancer (MIBC) who had both their FFPE tumor tissues and MRI images (before and after treatment) available for subsequent histological analysis and image analysis.
Cell lines and cell culture
The MB49, NIH/3T3, HEK293T, and RAW 264.7 cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS), penicillin, and streptomycin. All the cells were cultured at 37°C in a humidified incubator with 5% CO2. All cell lines were regularly tested and found to be negative for mycoplasma, and were authenticated by a short tandem repeat marker profiling approach (IGXN20178, IGE Biotechnology, Guangzhou, China).
For the experiments regarding the epigenetic regulation of the target gene, we used selective inhibitors of histone methyltransferase (EZH2) and demethylase (pan-KDM5) to test whether gene expression could be altered by DNA methylation mechanisms. Specifically, NIH/3T3 cells were cultured in a complete medium supplemented with 10 µmol/L methionine or were treated with 15 nmol/L EZH2 inhibitor PF-06726304 (Selleck, Houston, Texas, USA), 25 nmol/L pan-KDM5 inhibitor CPI-455 HCl (Selleck, Houston, Texas, USA) or vehicle (dimethyl sulfoxide, DMSO, MP Biomedicals, USA) for 72 hours.
Primary cell culture and treatment
Primary CAFs were isolated from bladder cancer samples obtained by cystectomy. Briefly, tissues were digested by collagenase type I, collagenase type III and hyaluronidase (1.5 mg/mL, Sigma Aldrich) at 37°C. Thereafter, primary CAFs were cultured in DMEM with 10% FBS. The CAFs from the first to fifth passages were used in experiments.
Bone marrow-derived macrophages (BMDMs) from mice were isolated by flushing the femurs and tibias with DMEM supplemented with 10% heat-inactivated FBS. After filtration through a 70 µm nylon cell strainer (BD Biosciences, BD Biosciences, San Jose, California, USA), the bone marrow cells were centrifuged once (5 min at 250×g, 4°C) and resuspended in 1×Red Blood Cell Lysis Buffer (CWBIO, Taizhou, Jiangsu, China) to lyse red blood cells. BMDMs were cultured in a complete medium containing 25 ng/mL macrophage colony-stimulating factor (m-CSF) (315–03, PeproTech, Rocky Hill, New Jersey, USA). On days 2 and 4, half of the medium was replaced with a complete medium containing m-CSF. On day 6, BMDMs were treated with a conditional medium for the experiment.
Colony formation assay
A co-culture experiment was performed by seeding MB49 cells (1×103) in the lower chamber and NIH/3T3 cells with stable overexpression of NNMT or vector controls (1×105) in the upper chamber of a 6-well transwell apparatus with a 0.4 µm pore size (Corning Incorporated, Corning, New York, USA), and the cells were cultured for 7 days. Subsequently, colonies were fixed with 4% paraformaldehyde and stained with crystal violet for 30 min. Colonies were counted and the counts were normalized to the average of each control group.
Transwell assays
The CAFs (1×106) were transduced with small Interfering RNA (siRNA) oligos provided by GenePharma (Shanghai, China) and Lipofectamine RNAiMAX (Life Technologies) and then cultured with fresh complete medium after 8 hours. Thereafter, 700 µL of the culture supernatant was placed into the lower chambers as a chemoattractant, and approximately 2×105 monocytes were suspended in 200 µL serum-free medium and added to the upper chambers of a 6-well transwell apparatus with 0.8 µm pore size (Corning Incorporated, Corning, New York, USA). After incubation for 8 hours, the cells in the upper chamber were gently removed with cotton swabs, and the cells on the lower surface were fixed with methanol and stained with 0.1% crystal violet for photographing and counting.
Antibodies
The NNMT was obtained from Santa Cruz Biotechnology (California, USA). CD206 and CD204 were purchased from Minneapolis (Minnesota, USA). Monoclonal anti-mouse antibodies for F4/80, arginase 1 (Arg1), platelet-derived growth factor receptor alpha (PDGFRα) were purchased from Cell Signaling Technology (Massachusetts, USA). The inducible nitric oxide synthase (iNOS) was obtained from Abcam (Cambridge, UK). The CD68 and cytokeratin were purchased from ZSGB-BIO (Beijing, China). The information on the antibodies is summarized in online supplemental table S1.
Supplemental material
Immunohistochemistry and evaluation
FFPE tissues from the SYSMH and SYSUCC cohorts were sliced into 4 µm sections, and immunohistochemical staining was performed according to previously described methods.26 Briefly, after deparaffinization and rehydration, the sections were treated with Tris-ethylenediaminetetraacetic acid buffer (pH 9.0), microwaved, and treated with 3% hydrogen peroxide for 10 min to quench endogenous peroxidase. The primary antibodies NNMT (1:100), CD206 (1:2000), F4/80 (1:1000), Arg1 (1:1000), CD204 (1:2000), and iNOS (1:500) were applied overnight in a humidified chamber at 4°C. After the primary antibodies were washed off, the sections were treated with the appropriate peroxidase-conjugated secondary antibodies for 30 min at 37°C, incubated with 3,3′-diaminobenzidine tetrahydrochloride in an Envision System (Dako, Glostrup, Denmark), and counterstained with hematoxylin (ZSGB-BIO, Beijing, China).
Immunoreactivity in each slide was evaluated independently by two researchers with experience in pathology blinded to the clinicopathological and survival data. NNMT expression was evaluated using the H-score according to the intensity and percentage of positive cells. This study focused on NNMT expression in the stroma. In a small subset of samples, some epithelial tumor cells also expressed NNMT, which was ignored. Staining intensity was classified as follows: 0 (no staining), 1+ (weak staining), 2+ (moderate staining), or 3+ (strong staining). The staining profile the samples is summarized in online supplemental table S2. The density of CD206+ macrophages in the tumor stroma was calculated as the mean count of the five most representative high-power fields (40× magnification) for each area of all specimens. The densities of F4/80+, Arg1+, CD204+, and iNOS+ macrophages in mouse tumor tissues were the mean counts of three representative high-power fields (20×magnification).
Immunofluorescence
Multiplexed fluorescent immunohistochemistry (IHC) was performed on FFPE tissues according to a previously described OPAL serial immunostaining protocol.27 Briefly, FFPE tissue sections were processed according to the standard IHC protocol described above. The primary antibodies used for staining were NNMT (1:100), PDGFRα (1:100), CD68 (1:100), and cytokeratin (1:100). Signal detection was performed using OPAL520, OPAL570, and OPAL620 (1:500, Perkin-Elmer, Waltham, Massachusetts, USA) in Tyramide Signal Amplification Plus working solution at room temperature for 10 min. Counterstaining was performed using 4',6-diamidino-2-phenylindole (DAPI, diluted 1:1000, 471224, Sigma-Aldrich, St. Louis, Missouri, USA) and subsequently mounted using Mowiol 4–88 medium (81381, Sigma-Aldrich, St. Louis, Missouri, USA). Additionally, one stained section was imaged at 40× magnification using a Zeiss LSM800 system (Zeiss, Jena, Germany).
RNA isolation and RT-qPCR analysis
Total RNA was extracted from cells and fresh-frozen tissue samples using the RNAiso Plus reagent (Takara, Kyoto, Japan) according to the manufacturer’s protocol. Complementary DNA (cDNA) was obtained from 1 µg of total RNA using the PrimeScript RT Master Mix (Takara, Kyoto, Japan). A 10 µL reaction volume containing 50 ng cDNA was used for RT-qPCR amplification in 384-well plates on a Roche LightCycler 480 II (Roche, Switzerland). The thermal cycling parameters included an initial denaturation at 94°C for 30 s, followed by 45 cycles at 94°C (25 s), 60°C (45 s), and 72°C (45 s). All experiments were performed independently at least three times. The 2−ΔΔCT method, using the average cycle threshold (CT) calculated from duplicate reactions for each tested sample, was used to determine the relative quantification of the target mRNA levels. The sequences of the PCR primers are listed in online supplemental table S3.
Western blot analysis
To determine the protein levels in cells, total protein was extracted using Radio Immunoprecipitation Assay lysis buffer (Beyotime Biotechnology, Shanghai, China) with a protease and phosphatase inhibitor cocktail (Beyotime Biotechnology, Shanghai, China), and the protein concentrations were measured with a bicinchoninic acid disodium salt hydrate Protein Assay Kit (CWBIO, Taizhou, Jiangsu, China). A 30 µg protein sample was subjected to sodium dodecyl sulfate–polyacrylamide gel electrophoresis at different densities and transferred to a polyvinylidene fluoride membrane (Millipore, Billerica, Massachusetts, USA). After blocking with 5% bovine serum albumin (Beyotime Biotechnology, Shanghai, China), the membranes were incubated with primary antibodies overnight at 4°C and then incubated with horseradish peroxidase-conjugated secondary antibodies (diluted 1:2000, Cell Signaling Technology, Danvers, Massachusetts,USA) for 1 hour at room temperature after washing three times with Tris-buffered saline with Tween. Signals were visualized using enhanced chemiluminescence detection reagents (Millipore, Billerica, Massachusetts, USA) and imaged using an iBright CL1000 Imaging System (Invitrogen, USA). All experiments were performed independently at least thrice.
NNMT plasmid transfection and stable cell strain selection
The pCDH/NNMT or pCDH/vector plasmids were purchased from IGE Biotechnology (Guangzhou, China). Lentivirus production and infection were performed according to the manufacturer’s protocol. NIH/3T3 cells were grown in a complete medium containing 2 mg/L puromycin (Beyotime Biotechnology, Shanghai, China) for 2 weeks, and cells with stable NNMT overexpression and vector controls were selected for further analysis. The plasmid used for Clustered Regularly Interspaced Short Palindromic Repeats-CRISPR-associated protein 9 (CRISPR‒Cas9)-mediated knockout (KO) of saa was produced following the protocol using the lentiCRISPR V.2 backbone.28 The short hairpin RNA (shrNA) or CRISPR vector was transfected into 293 T cells along with the psPAX2 and pMD2.G plasmids for viral packaging. Cells were infected with lentivirus and selected with puromycin.
Enzyme-linked immunosorbent assay
A total of 1.5×105 NIH/3T3 cells were grown to confluence in 6-well plates, and the supernatants were collected, centrifuged (5 min at 250×g), aliquoted, and stored at −80°C. The concentration of Saa3 was measured using an ELISA kit (MultiSciences/LIANKE, Biotech, Hangzhou, Zhejiang, China) according to the manufacturer’s instructions. Briefly, culture supernatants and standards were incubated in microtiter plates coated with the appropriate capture antibody at 4°C overnight. Following the application of the detection antibody, substrate tetramethyl benzene (TMB), and 2 N sulfuric acid to stop the reaction, the absorbance was read using an ELISA microtiter plate reader at a wavelength of 490 nm. Saa3 levels were corrected for each sample using standard curve-calculated chemokine standards.
Chromatin Immunoprecipitation assay
Chromatin immunoprecipitation assay (ChIP) assays were performed using a Pierce Magnetic ChIP Kit (Thermo Fisher Scientific, Waltham, Massachusetts, USA), according to the manufacturer’s protocol. Briefly, cells were fixed with 1% formaldehyde for 10 min at room temperature, and then 10× glycine was added to neutralize the excess formaldehyde. Following lysis with sodium dodecyl sulfate lysis buffer and ultrasonication to shear the DNA into 200–500 nt fragments, equal aliquots of chromatin supernatants were immunoprecipitated with anti-trimethyl-histone H3 Lys27 (H3K27me3, 9733, Cell Signaling Technology, Danvers, Massachusetts, USA) and immunoglobulin G (IgG, 3900, Cell Signaling Technology, Danvers, Massachusetts, USA) at 4°C overnight. RT-PCR was used to evaluate enrichment using the primers listed in online supplemental table S3.
Liquid chromatography tandem mass spectrometry (LC-MS/MS)
NIH/3T3 cells (2×106) were harvested and centrifuged at 1,400×g at 4°C. Protein samples precipitated with acetone-trichloroacetic acid (TCA) were digested with trypsin to generate proteolytic peptides that were labeled with iTRAQ reagents (Thermo Fisher Scientific, Waltham, Massachusetts, USA) and analyzed using liquid chromatography tandem mass spectrometry (LC-MS/MS) for both identification and quantification at the Bioinformatics and Omics Center, Sun Yat-sen Memorial Hospital, Sun Yat-sen University (Guangzhou, China).
To prepare samples for metabolite profiling, the cells were washed twice with ice-cold phosphate buffered saline (PBS), snap-frozen in liquid nitrogen, and mixed with a methanol/acetonitrile/H2O mixture. The samples were subjected to three freeze–thaw cycles and centrifuged at 12,000 g for 15 min at 4°C. Supernatants were collected as described above for ELISA and added to a methanol/acetonitrile/H2O mixture before centrifugation at 16,000×g for 10 min at 4°C to remove debris. The metabolite-containing supernatant was evaporated until dry and reconstituted in 100 µL formic acid. Targeted measurements of intracellular metabolites were performed at the Bioinformatics and Omics Center of Sun Yat-sen Memorial Hospital, Sun Yat-sen University (Guangzhou, China). The relative metabolite abundance was quantified using the integrated peak area for a given multiple-reaction-monitoring transition. Data are representative of three independent biological experiments, each containing three technical replicates for each condition.
Murine experiments
All C57BL/6 mice (4–5 weeks old) were purchased from the Experimental Animal Center, Sun Yat-sen University, and all animal experiments were conducted in compliance with the Institutional Animal Care and Use Committee of Sun Yat-sen University (Approval Number SYSU-IACUC-2020-B0206). Mice in each group were subcutaneously injected into the right flank with 5×104 MB49 mixed with 5×105 NIH/3T3 cells expressing the pCDH/vector or pCDH/Nnmt constructs in 50% Matrigel (BD Biosciences, San Jose, California, USA) diluted in serum-free Dulbecco’s modified Eagle’s medium.
For the NNMT inhibitor experiment, treatment with 10 mg/kg NNMT inhibitor 5-Amino-1-methylquinolinium iodide (#6900, Tocris Bioscience, Bristol, UK) in PBS or PBS (vehicle control) was initiated 24 hours after injection and administered intraperitoneally every 12 hours. Mice were treated with an anti-PD-L1 antibody (10 mg/kg, BE0101, Bio X Cell, West Lebanon, New Hampshire, USA) on day 5 and then treated once every 3 days. Tumor size was measured every 2 days. The tumor volume was calculated using the following equation: (length × width2)/2. The mean tumor volume of each group was used to plot tumor growth curves. The diameter of any one tumor did not exceed 2.0 cm in mice. On day 19, at the end of the experiment, the mice were euthanized by dislocation under anesthesia. The total tumor burden was dissected and divided into two parts for further analysis of tumor-infiltrating lymphocytes (TILs). One part of the tumor sample was fixed in 10% neutral-buffered formalin and embedded in paraffin blocks. Serial sections were used for IHC or immunofluorescence. The other part of each tumor sample was used for flow cytometry analysis.
Tissue digestion for cell isolation and FACS analysis
Tumors from the animal experiments were minced with a scalpel and digested with a collagenase mixture, containing collagenases (collagenase I, LS004194, 0.17 mg/mL, collagenase II, LS004174, 0.056 mg/mL, collagenase IV, LS004186, 0.17 mg/mL, Worthington Biochemical Corporation, Lakewood, New Jersey, USA), deoxyribonuclease l (LS002138, 0.025 mg/mL, Worthington Biochemical Corporation, USA), and elastase (LS002290, 0.025 mg/mL, Worthington Biochemical Corporation, Lakewood, New Jersey, USA). The digested substance was passed through a 70 µm filter (BD Biosciences, Franklin Lakes, New Jersey, USA) and incubated with red blood cell lysis buffer (CWBIO, Taizhou, Jiangsu, China) for 5 min to eliminate red blood cells.
For the fluorescence-activated cell sorting (FACS) assay, single-cell suspensions of tumor cells were used for FACS staining, according to standard protocols. Briefly, after using anti-mouse CD16/32 (Mouse BD Fc Block, 2.4G2, BD Biosciences, Franklin Lakes, New Jersey, USA) to avoid non-specific binding of antibodies to fc receptor gamma chain (FcRγ), the TILs were incubated with FACS buffer containing fluorochrome-conjugated antibodies or isotype control for 30 min at 4°C in the dark. The analysis was performed using a CytoFLEX Flow Cytometer (Miami, Florida, USA). Myeloid cells and macrophages were initially gated as large, single, or live cells. Myeloid cells were defined as the CD45+/CD11b+ cell population. Macrophages were designated as CD45+/CD11b+/F4/80+.
Data availability
The single-cell RNA-sequencing (scRNA-seq) data were downloaded from the EMBL-EBI datasets under BioProject PRJNA662018 (https://www.ebi.ac.uk/ena/browser/view/PRJNA662018?show=reads).11 Transcriptome data and clinical information of 408 patients with bladder cancer were downloaded from The Cancer Genome Atlas (TCGA) database using the R package TCGA biolinks.29 H3K27me3 ChIP-seq data provided by Professor Mark A Eckert were obtained from the Gene Expression Omnibus (GEO) (GSE124015).22 Immunotherapy data for metastatic urothelial cancer were obtained from the IMvigor210 cohort (http://research-pub.gene.com/IMvigor210CoreBiologies).14
Bioinformatics analysis of IMvigor210 dataset and scRNA-seq data
Differentially expressed genes (DEGs) associated with immunotherapy resistance were identified in the IMvigor210 cohort using the empirical Bayesian approach of the DEGseq2 R package. The criteria for determining differential DEGs were set with an adjusted p<0.05, and a |fold change|>2. Gene set enrichment analysis (GSEA) was performed on a matrix of all genes detected by the desktop tool to determine significant and concordant differences between a set of biological processes from the Molecular Signatures Database (MsigDB). The mRNA data of 408 patients with complete prognostic data from the TCGA-BLCA dataset were used to identify genes co-expressed with CAFs markers, including TGFB1, FAP, PDGFRα, and COL1A1. The strong co-expression criteria were set as Pearson correlation coefficients (r)>0.4 and p<0.05. The common genes co-expressed with markers of CAFs were determined using the VennDiagram R package.
For the scRNA-seq data, single-cell data from eight patients with UBC were obtained from the EMBL-EBI datasets and processed using Cell Ranger V.4.0.0 (https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/what-is-cell-ranger) to demultiplex cellular barcodes, map reads to the transcriptome, and downsample reads. Sequentially, the R package Seurat V.4.0.1 (https://satijalab.org/seurat/) was used for the standard processing of the unique molecular identifier (UMI) count matrix according to the guidelines. For data quality control, cells with fewer than 1000 UMIs, more than 6000 genes, or mitochondrial gene expression exceeding 10% were considered low-quality cells and were removed. Fast mutual nearest neighbor correction (FastMNN) (https://rdrr.io/github/LTLA/batchelor/man/fastMNN.html) was used to remove potential batch effects. Subsequently, principal component analysis was performed on an integrated data matrix with the top 50 principal components, and the FindClusters function was used to identify the main cell clusters with 0.3 resolution based on the ElbowPlot function. Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) plots were generated for visualization. We used conventional markers described in previous studies to identify biological cell types.11 15 30
Statistical analysis
31We used eight deconvolution methods, including (CIBERSORT, TIMER, xCell, MCPcounter, ESITMATE, EPIC, IPS, and quanTIseq) implemented in the Immuno-Oncology Biological Research (IOBR) package to predict the presence of infiltrating stromal/immune cells (cell fraction) in the TCGA-BLCA dataset.31 Bladder cancer subtypes based on RNA-seq data from TCGA dataset were classified using the consensusMIBC R package.32
For the analysis of NNMT gene expression, the optimal cut-off value of NNMT expression in TCGA-BLCA dataset was determined using the Survminer R package. For the IHC analysis, we used the median of the H-scores as the cut-off value to classify the high-expression and low-expression groups. Cumulative overall survival after surgery was estimated using the Kaplan-Meier method and compared using a log-rank test. In addition, to explore the association between variables and clinical outcomes, univariate analysis was performed, and the multivariate Cox proportional hazards model was used to estimate the adjusted HRs and 95% CIs to identify independent prognostic factors. Comparisons between groups were performed using a two-tailed Student’s t-test or one-way analysis of variance, followed by a post hoc test. Differences were considered statistically significant at p<0.05. Pearson’s χ2 test or Spearman’s correlation analysis was used to determine the correlation between the two variables, as appropriate.
Statistical analyses were conducted using R software (V.4.0.1, The R Foundation for Statistical Computing; http://www.r-project.org/), the SPSS software (V.22.0; SPSS, Chicago, Illinois, USA) and Prism 7.01 (GraphPad Software, La Jolla, California, USA), according to the data types. Experiments in which representative images are presented were performed with at least three biological replicates. Data are reported as the mean±SEM unless otherwise noted in the figure legends.
Results
NNMT was associated with immunotherapy resistance in UBC
To identify novel genes that may specifically play roles in immune resistance in UBCs, using the clinical trial dataset IMvigor210, an RNA-Seq dataset from a phase II trial of atezolizumab (MPDL3280A) in metastatic urothelial carcinoma,33 we investigated the role of CAFs in conferring resistance to PD-1/PD-L1 blockade. We focused on patients with the immune-excluded phenotype (n=113), in which CD8+ T cells were spatially separated from tumor cells in UBC tissues and showed a significant association with a poor response to ICIs.14 As a result, we identified 901 genes that were significantly dysregulated in 85 anti- PD-1/PD-L1 non-responders compared with 28 responders (figure 1A). Importantly, fibroblast-related genes were ranked at the top of the hits among the dysregulated genes in the non-responders (online supplemental figure S1A). Furthermore, GSEA showed that fibroblast-related pathways were significantly enriched in the non-responders (online supplemental figure S1B).
Supplemental material
NNMT+ CAFs is associated with immunotherapy resistance and an independent unfavorable prognostic marker in UBC. (A) An MA plot of 113 excluded phenotype patients demonstrated 901 significant differentially expressed genes (DEGs, |fold change|>2.0, adjusted p<0.05) following non-responders (n=85) compared with responders (n=28) from the IMvigor210 dataset. (B) Venn diagram details the number of genes co-expressed (r>0.4, p<0.05) with inflammation related fibroblast makers (TGFB1, FAP, PDGFRα and COL1A1) from the urothelial carcinoma of bladder TCGA database. (C) Venn diagram showing DEGs from (A) and genes co-expressed with inflammation related fibroblast makers (TGFB1, FAP, PDGFRα and COL1A1, top). Α heatmap of correlation of overlapping genes between with fibroblast makers from the TCGA dataset (n=408, bottom). (D) Representative images of pre- (left) or post- (right) treatment stromal NNMT expression and MRI from two patients with UBC who treated with four courses of neoadjuvant tislelizumab combined with cisplatin and gemcitabine (ICIs+GC chemotherapy). Scale bars, 100 µm (magnification, 20×). The yellow arrow indicates the tumor and the white dotted line marked the tumor lesion field. (E) Change in tumor lesion length was apparent in 13 patients with UBC who were treated with four courses of neoadjuvant tislelizumab combined with cisplatin and gemcitabine (ICIs+GC chemotherapy). Scale bars, 100 µm (magnification, 20×). Blue lines, NNMT low expression group; red lines, NNMT high expression group. (F) Paraffin-embedded UBC sections were subjected to multiplex immunofluorescence staining for NNMT (green), PDGFRα (red) and CK (white). Scale bars, 20 µm (magnification, 40×). (G) UMAP plot of single cells profiled in the present study colored by major cell types (left). UMAP plot indicating the expression level of NNMT in the fibroblast subpopulation (right). ScRNA-seq data were obtained from the EMBL-EBI datasets (PRJNA662018). (H) Cumulative overall survival curves of the patients in the SYSMH cohort (left) and the SYSUCC cohort (right). Patients from the SYSMH cohort and the SYSUCC cohort were divided into two groups according to the median IHC score of stromal NNMT. Kaplan-Meier survival estimates and log-rank tests were used to analyze the prognostic significance of the two groups. (I) Cumulative overall survival curves of the UBC patients in the TCGA database. Patients from TCGA were divided into two groups according to the optimal cut-off value of mRNA NNMT that was identified using the Survminer R package. Kaplan–Meier survival estimates and log-rank tests were used to analyze the prognostic significance of the two groups. BLCA, bladder urothelial carcinoma; CAFs, cancer-associated fibroblasts; GC, gemcitabine plus cisplatin chemotherapy; ICIs, immune checkpoint inhibitors; IHC, immunohistochemistry; MA, logged intensity ratio (M) versus mean logged intensities (A); mRNA, messenger RNA; NNMT, nicotinamide N-methyltransferase; scRNA-seq, single-cell RNA-sequencing; SYSMH, Sun Yat-sen Memorial Hospital; SYSUCC, Sun Yat-sen University Cancer Center; TCGA, The Cancer Genome Atlas; UBC, urothelial bladder cancer; UMAP, Uniform Manifold Approximation and Projection.
Next, to interrogate specific CAF genes that drive immune resistance and tumor progression in UBC, we conducted a co-expression analysis in TCGA- RNA-Seq data of UBC and identified genes strongly co-expressed with the well-established inflammatory CAFs markers TGFB1, FAP, PDGFRα, and COL1A1.11 34. Consequently, we obtained 77 genes strongly co-expressed with the inflammatory CAFs markers (r>0.4, p<0.05, figure 1B). Then, we intersected the co-expressed gene set with the dysregulated genes in non-responders using Venn diagrams, resulting in 10 fibroblast-co-expressed genes that were differentially upregulated in non-responders (figure 1C). Among these genes, NNMT, a key enzyme in NAD+ metabolism, is a novel target that is significantly associated with immune resistance in UBCs (online supplemental figure S1C).
To determine the association between NNMT expression and the clinical efficacy of ICIs in UBC, we assessed the expression levels of NNMT in the stromal cells of 13 patients with UBC (cohort D) who were treated with four courses of neoadjuvant tislelizumab combined with gemcitabine and cisplatin (GC chemotherapy). Based on the criteria of Response Evaluation Criteria in Solid Tumors (RECIST V.1.1), we found that the tumor sizes of eight patients with low NNMT expression in pretreatment tissues were significantly reduced, suggesting a favorable response to neoadjuvant treatment. In contrast, the other five patients with high expression levels of NNMT showed no response to this neoadjuvant treatment, and their tumor size increased or remained unchanged (figure 1D–E). These data suggest that elevated NNMT expression in the stroma is significantly associated with resistance to immunotherapy in UBC.
NNMT showed dominant expression on cancer-associated fibroblasts in UBC
To characterize the role of NNMT in UBCs, we examined NNMT expression using an IHC assay in samples from the SYSMH cohort (cohort A) and found that NNMT demonstrated stromal expression, specifically within fibroblasts in UBC samples (the first slide in the upper panel). Unlike renal cancer and colon cancer (see the slides in the lower panel), where NNMT expression was predominantly observed in tumor cells (online supplemental figure S2A). Furthermore, we confirmed that NNMT was co-expressed with the CAF marker PDGFRα by multiplex immunofluorescence analysis of UBC tissues (figure 1F), consistent with the results of the re-analysis of scRNA-seq data of eight human UBC samples11 (figure 1G). Co-expression analysis of the mRNA data from 32 patients with UBC and TCGA-BLCA dataset (n=408) further supported the finding that NNMT was mainly expressed in CAFs in UBC (online supplemental figure S2B,C). In addition, a higher proportion of fibroblasts was observed in samples with higher NNMT expression (the cut-off value was determined using the Survminer R package) (all p<0.001; online supplemental figure S2D).
Taken together, these data indicate that the NNMT protein is mainly expressed in the stromal compartment of UBC rather than in the cancerous epithelial compartment. Therefore, we focused on NNMT in the stromal compartments of patients with UBC and defined these subpopulations as NNMT+ CAFs.
NNMT+ CAFs was an independent unfavorable prognostic marker in UBC
To determine the clinical value of NNMT+ CAFs in UBC, we used FFPE tissue specimens from two cohorts (SYSMH and SYSUCC) for subsequent clinical analyses. The characteristics of the study participants are summarized in table 1. We assessed the IHC scores of NNMT in stromal cells from 337 cases from the two centers (online supplemental table S2) according to our criteria. The H-score of stromal NNMT expression was calculated using the methods described in Methods according to a previous study.35 Subsequently, the tissues from the SYSMH and SYSUCC cohorts were classified into high stromal NNMT and low stromal NNMT groups according to the H-score. Kaplan-Meier analysis showed a significant association between the high H-score of stromal NNMT and poor overall survival (OS) (all p<0.001 in the SYSMH and SYSUCC cohorts; figure 1H). A similar result was observed for NNMT mRNA expression in the TCGA-BLCA dataset (p=0.033; figure 1I).
Clinicopathological characteristics of patients with UBC in the SYSMH cohort and the SYSUCC cohort
Additionally, high stromal NNMT expression in UBC was a significant independent adverse predictor of OS in the multivariate Cox regression analysis (HR=2.437 (95% CI=1.069 to 5.554), p=0.034 in the SYSMH cohort; HR=4.904 (95% CI=2.650 to 9.074), p<0.001 in the SYSUCC cohort; table 2). Consistently, we found that elevated expression levels of stromal NNMT were significantly associated with clinical characteristics, such as high T stage, positive lymph node status, and high grade (online supplemental table S4; online supplemental figure S3B-D). A similar relationship was found after stratification by tumor stage (online supplemental figure S3E,F) and histological grade (online supplemental figure S3G,H). Together, these data demonstrate that NNMT+ CAFs are significantly associated with an unfavorable UBC prognosis.
Multivariate analysis of the factors associated with overall survival in the SYSMH cohort and the SYSUCC cohort
NNMT inhibitor enhanced the antitumor effects of anti-PD-L1 antibody treatment and is associated with decreased recruitment of macrophages
To determine the function of NNMT and whether NNMT inhibitor (NNMTi) treatment can therapeutically enhance the efficacy of ICI-based immunotherapy, we used MB49, a common-used murine bladder carcinoma cell line for animal models. Since the expression levels of NNMT is low in parental MB49 tumor cells, to better mimic the dynamic cell-cell interactions to study the effects of NNMT+ CAFs contributing to tumor progression in UBC, following the approach using in a previous study,22 we conducted co-injection of the mixed MB49 tumor cells with Nnmt-overexpressing NIH/3T3 cells into mice to establish tumor models, and then treated with a combination of anti-mouse-PD-L1 antibody and NNMTi (5-Amino-1-methylquinolinium iodide; figure 2A). We observed that, compared with the control group, the NNMT inhibitor decreased the tumor burden (figure 2B–E), increased necrosis, and decreased proliferation (represented by Ki-67 staining) (figure 2F; online supplemental figure S4A,B). Notably, the NNMT inhibitor increased the antitumor efficacy of the anti-PD-L1 antibody treatment, in which we observed a clear trend of decrease in tumor growth and increase in tumor necrosis compared with the group treated with anti-mouse-PD-L1 antibody alone in the MB49 model (figure 2B–E).
NNMTi treatment reduces mouse bladder tumor via regulating macrophages infiltration and enhancing the antitumor effects of anti-PD-L1 antibody. (A) Treatment protocol. Mouse bladder cancer cells (MB49) were co-injected with NIH/3T3 fibroblasts expressing NNMT constructs in C57BL/6 mice, which were divided into four groups (n=6 mice per group). One group received PBS as a control; the other groups received NNMTi (5-Amino-1-methylquinolinium iodide, 10 mg/kg, two times a day) or/and anti-PD-L1 (10 mg/kg, every 3 days). (B) Representative figure of tumor burden of the four groups from (A). (C) Histogram analysis demonstrating the weight of tumors after surgical dissection (left) and the volume of tumors measured every 2 days (right). Data were presented as the mean±SEM for the indicated number of mice per treatment (n=6). Statistical significance was determined by one-way ANOVA. (D) Representative figure of tumor burden of the four groups (n=10 mice per group). (E) Histogram analysis demonstrating the weight of tumors after surgical dissection (left) and the volume of tumors measured every 2 days (right). Data were presented as the mean±SEM for the indicated number of mice per treatment (n=10). Statistical significance was determined by one-way ANOVA. (F) Representative figure of IHC staining of macrophage markers (F4/80, CD204, Arg-1 and iNOS) in mouse subcutaneous tumor tissue from PBS (Group 1) or NNMTi (Group 2) (n=6 per group). Data were shown as the mean±SEM of values for six tumors. Two-tailed Student’s t-tests were used to evaluate the statistical significance of differences between two groups. *p<0.05, **p<0.01, and ***p<0.001. ANOVA, analysis of variance; IHC, immunohistochemistry; NNMT, nicotinamide N-methyltransferase; NNMTi, nicotinamide N-methyltransferase inhibitor; PBS, phosphate-buffered saline; PD-L1, programmed death-ligand 1.
More importantly, we found that NNMTi treatment significantly enhanced the recruitment of pro-tumor macrophages (figure 2F), including decrease in F4/80+ macrophages (234.0±12.82 vs 122.8±8.30, p<0.001), CD204+ macrophages (259.8±9.62 vs 133.8±18.69, p=0.001), and Arg-1+ macrophages (171.4±22.53 vs 59.5±8.40, p<0.001). However, there was a significant increase in antitumor iNOS+ macrophages (10.78±1.20 vs 35.94±2.47, p<0.001; figure 2F). These data indicated that NNMT expression is associated with macrophage recruitment. The use of NNMT inhibitors with anti-PD-L1 antibodies may improve the therapeutic efficacy of ICI-based immunotherapy in UBC.
NNMT+ CAFs was associated with inflammatory CAF phenotypes and the aggressive UBC subtypes
To characterize NNMT expression profiles in the CAF subpopulations, we re-analyzed the UBC scRNA-seq raw data (EMBL-EBI dataset PRJNA662018). We first conducted clustering analysis of CAF subpopulations in single-cell RNAseq data. We annotated the cell types based on CAFs markers and found that CAFs in UBC could be divided into two distinct subpopulations, namely, mCAF (myofibroblast CAFs) and iCAF (inflammatory CAFs) (online supplemental figure S5A,B). GSEA showed that the upregulated genes in mCAFs were enriched in muscle contraction, whereas the upregulated genes in iCAFs were enriched in TGF-β signaling, extracellular matrix degradation, and the cytokine–cytokine receptor interaction pathway (online supplemental figure S5C). These data suggest that NNMT+ CAFs are linked to the inflammatory CAF phenotypes.
Importantly, NNMT, a key enzyme in NAD+ metabolism and biological oxidation pathways, was expressed at higher levels in iCAFs than in mCAFs in UBC (p<0.001; online supplemental figure S5D). Further, we evaluated the activation of NAD+ metabolic pathways in different cellular compartments of UBC (online supplemental figure S5E). Surprisingly, we found that NAD+ biosynthesis and biological oxidation pathways were significantly enriched in iCAFs than in mCAFs (online supplemental figure S5F).
Next, we investigated the expression profiles of NNMT in specific molecular subtypes of UBCs (online supplemental file 1), we found that NNMT expression was higher in non-Luminal-papillary subtypes such as the Luminal subtype, the Basal-squamous subtype, and the Luminal-infiltrated subtype, as compared with the luminal-papillary subtype (all p<0.001; online supplemental figure S5), suggesting that the non-Luminal-papillary subtypes were strongly associated with NNMT+ CAFs. Notably, it has been well-documented that non-Luminal-papillary subtypes are associated with stroma enrichment and lymphocyte infiltration. Consistently, NNMT expression was positively correlated with immune expression and stromal scores (r=0.67 and 0.86, all p<0.001, respectively; online supplemental figure S5J). Collectively, these results indicate that in UBC, NNMT is specifically expressed in cancer-associated fibroblasts, particularly in iCAFs. NNMT+ CAFs may contribute to tumor aggressiveness by regulating the tumor immune microenvironment.
NNMT+ fibroblasts were correlated with macrophage infiltration in patients
To test the hypothesis that NNMT+ CAFs were involved in the regulation of immune cell recruitment and function, we investigated the correlation between NNMT expression and immune scores using the CIBERSORT method on the TCGA-BLCA dataset. The results demonstrated that NNMT expression was significantly correlated with the M2 macrophage score, ranking at the top of the 22 immune cell types (r=0.33, p<0.001; online supplemental figure S6A). In addition, the macrophage scores were higher in the high NNMT group than in the low NNMT group (figure 3A; online supplemental figure S6B), and co-expression analysis of the mRNA data from 32 patients with UBC and the TCGA-BLCA dataset (n=408) showed that elevated expression levels of NNMT were significantly correlated with those of many TAM marker genes (online supplemental figure S6C,D).
NNMT+ fibroblasts are associated with macrophages infiltration. (A) A heatmap showing the scores of macrophages subtypes in the NNMT high and low groups detected by various methods using the IOBR package from the TCGA-BLCA cohort. (B) Representative pictures showing the colocalization of NNMT+ cells (green) and macrophage markers (CD68, red) in the stromal (ST) regions of UBC tumor tissue with low (top) or high (bottom) NNMT expression. Scale bars, 100 µm (magnification, 10×). (C) Representative images of migrated monocytes recruited after knockdown of NNMT in cancer-associated fibroblasts isolated from fresh human tumor tissues compared with the nonspecific control (n=3 per group, magnification, 20×). (D) Histogram analysis showing the number of migrated monocytes. All the data were shown as the mean±SEM of values from three independent experiments. (E) The efficiency of NNMT knockdown by two siRNAs compared with the non-specific control in CAFs was verified by RT-qPCR. (F–G) Consecutive sections were used for the immunohistochemical identification of stromal NNMT+ cells and CD206+ cells in UBC tissues. (F) Representative images of NNMT+ (top) and CD206+ (bottom) cells in tumor tissue with either low (left) or high (right) NNMT expression are shown. Scale bar, 100 µm (magnification, 20×). (G) Positive association between NNMT+ cells and CD206+ cells in UBC tissues (n=117). NNMT_Si#1 and NNMT_Si#2 indicating two small RNA interfering agents for NNMT. Two-tailed Student’s t-tests were used to evaluate the statistical significance of differences between two groups. ***p<0.001. BLCA, bladder urothelial carcinoma; CAFs, cancer-associated fibroblasts; DAPI, 4',6-diamidino-2-phenylindole; IOBR, Immuno-Oncology Biological Research; NC, negative control; NNMT, nicotinamide N-methyltransferase; si, small interfering; TAM, tumor-associated macrophage; RT-qPCR, quantitative reverse transcription PCR; TCGA, The Cancer Genome Atlas; UBC, urothelial bladder cancer.
Furthermore, using dual-color immunofluorescence analysis, we observed that the expression levels of the marker for macrophages (CD68) were significantly correlated with the expression levels of NNMT in fibroblasts in UBC samples. Notably, we confirmed that NNMT+ fibroblasts frequently colocalized with stromal CD68+ macrophages in UBC tissues (figure 3B). Furthermore, to illustrate the geographical pattern of macrophages neighboring CAFs, we performed bioinformatics analyses in a publicly available spatial transcriptomics data of four UBC samples (GSE171351). We found that, in the NNMT high-expression spots, CD163+ macrophages were significantly spatially associated with elevated NNMT expression in fibroblasts, as compared with the NNMT low-expression spots, indicating the co-localization of NNMT+ CAFs and specific macrophages (CD163+) in UBC samples (online supplemental figure S7).
To confirm the macrophage recruitment function of NNMT in human fibroblasts, NNMT expression was knocked down in CAFs isolated from UBC samples, and treatment with conditioned medium from these cells significantly suppressed macrophage recruitment (all p<0.001; figure 3C–E). Consistently, we found that tissues with a high proportion of NNMT+ fibroblasts were enriched with CD206+ macrophages in the same histological areas (62.9±3.77 vs 39.3±3.34, p<0.001; figure 3F–G). These results suggest that NNMT+ fibroblasts play a vital role in the regulation within the same stromal area.
Nnmt+ NIH/3T3 cells promoted tumor progression by recruiting pro-tumor macrophages
To obtain the mechanistic insights into the effects of NNMT+ fibroblasts on macrophage function. We first analyzed whether NNMT+ fibroblasts were sufficient to induce macrophage recruitment and pro-tumor polarization. The supernatant of Nnmt-overexpressing NIH/3T3 cells was found to recruit more macrophages (RAW 264.7) and isolated BMDMs than vector controls according to a transwell assay (all p<0.001; figure 4A–B); indicating that Nnmt+ NIH/3T3 cells markedly enhanced the migration of macrophages. On the contrary, an NNMT inhibitor partly attenuated the recruitment of macrophages induced by Nnmt-overexpressing NIH/3T3 cells in vitro (all p<0.001; figure 4A–B). Moreover, the supernatant of Nnmt-overexpressing NIH/3T3 cells strongly activated the mitogen-activated protein kinase (MAPK and nuclear Factor Kappa-light-chain-enhancer of Activated B cells (NF-κB) pathways in RAW 264.7 cells (figure 4C) and polarized macrophages, which are involved in the upregulation of pro-tumor genes, such as Arg-1 (p<0.001), Il-6 (p=0.049), Pu.1 (p<0.001), and Cd206 (p<0.001) (online supplemental figure S8A) and the downregulation of antitumoral macrophage-related factors, such as Il-12α (p=0.006) and Cd86 (p=0.017) (online supplemental figure S8B).
Nnmt+ NIH/3T3 cells promoted tumor progression by recruiting pro-tumor macrophages. (A) Representative images of migrated mouse macrophages BMDMs (left) and RAW246.7 cells (right) were recruited by supernatants from Nnmt-overexpressing NIH/3T3 cells or pCDH and Nnmt-overexpressing NIH/3T3 cells treated with NNMTi (25 mg/mL) in six independent experiments (magnification, 20×). (B) Histogram analysis showing the number of migrated BMDMs (top) and RAW246.7 cells (bottom). The results are shown as the mean±SEM from six independent experiments. Two-tailed Student’s t-tests were used to evaluate the statistical significance of differences between two groups. (C) Western blotting was used to analyze the MAPK (phosphorylated p38, p44/42 and Akt) and NF-kB (phosphorylated p65) pathways in mouse macrophages (RAW246.7) treated with supernatants from Nnmt-overexpressing NIH/3T3 cells or vector. (D) Representative figure of the tumor burden of mouse bladder cancer cells (MB49) co-injected with NIH/3T3 fibroblasts expressing vector or Nnmt overexpression constructs in C57BL/six mice (n=5 per group). (E) Histogram analysis showing the weight of tumors after surgical dissection (left) and the volume of tumors measured every 2 days (right). The results are presented as the mean±SEM from five tumors. (F) Flow cytometric analysis of monocytes which were isolated fresh mouse subcutaneous tumor tissue from (D). The results are presented as the mean±SEM of values from five tumors. (G) Histogram analysis showing the number of infiltrated macrophages (F4/80+, CD204+, Arg-1+ and iNOS+ macrophages) in mouse subcutaneous tumor tissue from (D). Scale bar, 100 µm (magnification, 20×). Data were shown as the mean±SEM of values from five tumors (bottom). Two-tailed Student’s t-tests were used to evaluate the statistical significance of differences between two groups. *p<0.05, **p<0.01, and ***p<0.001. BMDM, bone marrow-derived macrophage; NF-κB, nuclear factor kappa-light-chain-enhancer of activated B cells; NNMTi, nicotinamide N-methyltransferase inhibitor; MAPK, mitogen-activated protein kinase.
Furthermore, to evaluate the effects of Nnmt+ fibroblasts on carcinogenesis, we assessed the proliferation of NNMT+ fibroblasts using colony formation assay. The results suggested that the overexpression of Nnmt did not change the proliferation of NIH/3T3 cells in vitro (online supplemental figure S8C). Meanwhile, the stable overexpression of Nnmt in NIH/3T3 cells did not directly affect the proliferation of mouse bladder cancer cells (MB49) in vitro (online supplemental figure S8D). The observations indicate that the pro-tumor function of NNMT+ fibroblasts might be mediated through immune cells in the TME.
To elucidate the mechanisms by which Nnmt-overexpressing NIH/3T3 cells facilitate tumor growth, we analyzed the functional roles of stable overexpression of Nnmt in the tumor growth of NIH/3T3 cells in vivo. We co-injected MB49 cells with NIH/3T3 cells stably overexpressing Nnmt or vector controls and found that the stable overexpression of Nnmt enhanced overall tumor burden (p=0.047) and tumor cell proliferation (p=0.032; figure 4D–E) in vivo. In addition, we analyzed the infiltration of tumor tissue by immunostaining and FACS with a gating strategy (online supplemental figure S8E). In agreement with the above results in vitro, FACS analysis demonstrated that the TME in the Nnmt-overexpressing NIH/3T3 cells group showed significantly greater infiltration of CD11b+F4/80+ macrophages than the vector control group (p=0.007 figure 4F); these changes were accompanied by enhanced recruitment of pro-tumor macrophages F4/80+ (106.6±9.30 vs 188.6±21.45, p=0.008), CD204+(140.8±10.39 vs 260.8±12.88, p<0.001), and Arg-1+ (58.8±7.08 vs 155.2±27.45, p=0.009) macrophages (figure 4G, online supplemental figure S8F), and downregulated antitumor iNOS+ macrophages (24.2±2.35 vs 12.4±3.87, p=0.031; figure 4G, online supplemental figure S8F) according to IHC analysis. Together, these results indicate that NNMT+ fibroblasts are important effectors that mediate macrophage reprogramming to promote tumor progression.
Nnmt+ NIH/3T3 cells recruited and reprogrammed macrophages via Saa3 signaling
To explore the molecular mechanisms by which Nnmt-overexpressing NIH/3T3 cells promoted the recruitment and polarization of pro-tumor macrophages, by using next-generation sequencing, we profiled the gene expression in the Nnmt-overexpressing NIH/3T3 cells, treated with/without an NNMTi (figure 5A). The results showed that 10 genes were upregulated in the Nnmt overexpression group compared with the vector control, which were concurrently downregulated in the NNMTi treatment group compared with the control (figure 5B). Among these, we found that Saa3, an essential molecule in the macrophage-mediated inflammatory process, was one of the most differentially expressed genes, suggesting that Saa3 might be directly regulated by Nnmt. Expectedly, we confirmed that the expression of Saa3 increased in Nnmt-transduced cells and decreased in NNMTi-treated cells using ELISA (all p<0.001; figure 5C) and western blotting (figure 5D).
Nnmt+ NIH/3T3 recruited and reprogrammed macrophages via Saa3 signaling. (A) Α heatmap of triplicate experiments demonstrating the top 250 significant differentially expressed genes following overexpression of Nnmt versus control in NIH/3T3 cells (|fold change|>2.0, adjusted p<0.05, left) and following NIH/3T3-overexpressing NNMT cells treated with NNMTi (25 mg/mL) versus PBS (|fold change|>1.5, adjusted p<0.05, right). Each column represents the indicated sample; each row indicates one mRNA. The red color scale represented a higher expression level, and the blue color scale represented a lower expression level. (B) Venn diagram detailed genes differentially expressed in Nnmt-overexpressing NIH/3T3 cells compared with those overexpressing the vector control and treatment with an NNMT inhibitor compared with PBS. Ten genes were found to be significantly upregulated in NIH/3T3 cell overexpressing Nnmt and attenuated with an NNMT inhibitor. Five genes were significantly decreased in Nnmt-overexpressing NIH/3T3 cells and increased in NNMTi-treated cells. (C–D) ELISA and western blotting verification of Saa3 expression, which was regulated by Nnmt in NIH/3T3 cells and attenuated by NNMTi treatment. The fold change compared with the average value of the vector group is presented as the mean±SEM. (E–F) ELISA and western blotting verification of Saa3 knockout efficiency in Nnmt-overexpressing NIH/3T3 cells by CRISPR–Cas9. The fold change compared with the average value of the vector group is presented as the mean±SEM from four independent experiments. (G) Representative images of migrated mouse macrophages, including BMDMs (left) and RAW246.7 cells (right) that were recruited by supernatants from Nnmt-overexpressing, Saa3 knockout or CRISPR–Cas9 vector control NIH/3T3 cells (n=9 per group; magnification, 20×). (H) Histogram analysis showing the number of migrated BMDMs (top) and RAW246.7 cells (bottom). The data are shown as the mean±SEM of values from nine independent experiments. Two-tailed Student’s t-tests were used to evaluate the statistical significance of differences between two groups. (I) Western blotting was used to analyze the MAPK pathway in RAW246.7 cells treated with supernatants from Nnmt-overexpressing NIH/3T3 cells and knocking out Saa3 or CRISPR–Cas9 vector. Phosphorylated Akt, p38, p44/42 and Sapk/Jun were examined. *p<0.05, **p<0.01 and ***p<0.001. BMDM, bone marrow-derived macrophage; CRISPR‒Cas9, Clustered regularly interspaced palindromic repeats (CRISPR) and CRISPR-associated (Cas) protein nine system for gene editing; MAPK, Mitogen-activated protein kinase; NNMTi, nicotinamide N-methyltransferase inhibitor; PBS, phosphate buffered saline.
To further investigate whether Nnmt-overexpressing NIH/3T3 cells recruit and polarize pro-tumor macrophages in a Saa3-dependent manner, we performed CRISPR–Cas9 KO of Saa3 in Nnmt-overexpressing NIH/3T3 cells. We first validated the efficiency of Saa3 KO using ELISA (all p<0.001; figure 5E) and western blot analysis (figure 5F). Moreover, rescue experiments indicated that KO of Saa3 in Nnmt-overexpressing NIH/3T3 cells reduced the recruitment of macrophages (RAW 264.7, cells and isolated BMDMs) as compared with that in the vector control group according to transwell assays (all p<0.001; figure 5G–H). In addition, we found that KO of Saa3 in Nnmt-overexpressing NIH/3T3 cells reduced MAPK pathway activation (figure 5I); downregulated pro-tumor macrophage genes (online supplemental figure S8G), such as Arg-1 (all p<0.001), Il-6 (p=0.005 and 0.003), Pu.1 (p=0.008 and p<0.001), and Cd206 (p<0.001 and 0.001); and upregulated antitumoral macrophage-related factors (online supplemental figure S8H), such as Il-12α (p=0.021 and 0.004) and Cd86 (p=0.005 and 0.005, respectively). Thus, these results indicate that Nnmt+ NIH/3T3 cells recruit and reprogram pro-tumor macrophages by upregulating Saa3 expression in vitro.
Knockout of Saa3 attenuated the effects of Nnmt+ NIH/3T3 cells on carcinogenesis in vivo
To confirm the roles of Saa3 in NIH/3T3 cells in vivo, MB49 cells with Saa3 KO or control Nnmt-overexpressing NIH/3T3 cells were co-injected into mice. As a result, the KO of Saa3 in Nnmt-overexpressing NIH/3T3 cells partly abolished the enhancement of the overall tumor burden (p=0.002 and 0.006, respectively) and tumor cell proliferation (all p=0.001) induced by Nnmt overexpression in vivo (figure 6A–B). The FACS analysis showed that the proportion of CD11b+F4/80+ macrophages infiltrating the TME was lower in the Saa3 KO group than in the control group (p=0.002 and p=0.080; figure 6C). The number of polarized pro-tumor macrophages was reduced in Saa3 KO Nnmt-overexpressing NIH/3T3 cells (figure 6D–E), which exhibited upregulated expression of F4/80+ (all p<0.001), CD204+ (all p<0.001) and Arg-1+ (all p<0.001) macrophages and downregulated expression of iNOS+ macrophages (p=0.004 and 0.009, respectively). These results suggest that Saa3 may play an important role in NNMT+ fibroblast recruitment and the reprogramming of pro-tumor macrophage polarization.
Knockout of Saa3 attenuates the effects of Nnmt+ NIH/3T3 cells on carcinogenesis in vivo. (A) Representative figure of the tumor burden of mouse bladder cancer cells (MB49) co-injected with Nnmt-overexpressing NIH/3T3 cells that was simultaneously transfected Saa3 CRISPR–Cas9 knockout or vector plasmid (n=5 per group). (B) Histogram analysis showing the weight of tumors after surgical dissection (left) and the volume of tumors measured every 2 days (right). The results were presented as the mean±SEM of values for five tumors. (C) Flow cytometric analysis of monocytes with isolated fresh mouse subcutaneous tumor tissue from (A). The results were presented as the mean±SEM of values from five tumors. (D–E) Representative figure of IHC staining of macrophage markers (F4/80, CD204, Arg-1 and iNOS) in mouse subcutaneous tumor tissue from (A). Scale bar, 100 µm (magnification, 20×). (E) Histogram analysis showing the number of infiltrated macrophages (F4/80+, CD204+, Arg-1+ and iNOS+ macrophage) in mouse subcutaneous tumor tissue. The data were shown as the mean±SEM of values for five tumors. Cas9 Saa3 #1(or sg#1) and Cas9 Saa3 #2(or sg#2) indicated two single guide RNA used for Saa3 knockout. Two-tailed Student’s t-tests were used to evaluate the statistical significance of differences between two groups. *p<0.05, **p<0.01 and ***p<0.001. CRISPR‒Cas9, Clustered regularly interspaced palindromic repeats (CRISPR) and CRISPR-associated (Cas) protein nine system for gene editing; IHC, immunohistochemistry; NC, negative control; NNMTi, nicotinamide N-methyltransferase inhibitor; PBS, phosphate buffered saline.
Saa3 was regulated by Nnmt expression by modifying histones methylation
Previous studies have revealed that, by attenuating the SAM/S-adenosyl-L-homocysteine ratio (SAM:SAH), NNMT could change the hypomethylation of DNA, RNA, or histones to drive gene expression.21 To investigate whether or not Nnmt could modify the methylation sites of Saa3 to up-regulates its expression, we conducted immunostaining analysis and targeted LC–MS experiment. We found that NIH/3T3 cells overexpressing Nnmt showed a decreased H3K27 trimethylation (figure 7A), decreased SAM levels, and increased SAH and 1-methylnicotinamide levels (online supplemental figure S9A). Consistently, our re-analysis of the H3K27me3 Chromatin immunoprecipitation (ChIP)-Seq data (GEO dataset GSE124015),22 showed that the H3K27me3 peaks on the promoter sequences of Saa3 were decreased in Nnmt-overexpressing NIH/3T3 cells (online supplemental figure S9B); and we confirmed in our cell models by ChIP-qPCR (p<0.001; figure 7B).
Saa3, the functional human ortholog of SAA1, regulated by Nnmt expression by modifying H3K27 methylation. (A) Western blot analysis was used to detect the expression of target proteins regulated in Nnmt-overexpressing NIH/3T3 cells compared with vector-transfected NIH/3T3 cells. (B) The promoters of target genes of Saa3 were detected by ChIP RT-qPCR in NIH/3T3 cells after overexpressing Nnmt (n=3 per group). IgG was used as a negative control. The values are normalized to the input and presented as the mean±SEM. (C) ELISA and (D) western blotting verification of Saa3, which was upregulated by Nnmt in NIH/3T3 cells and attenuated by the addition of methionine (10 µM) (n=4 per group). The results are presented as the mean±SEM of fold change for each sample with respect to the average value of the control group. (E–F) The positive correlation between the expression levels of NNMT and SAA1 by RT-qPCR in UBC tissues (n=32), (E) and in tumor samples from the TCGA-BLCA dataset (n=408), (F). (G) Representative images of migrated monocytes recruited by recombinant SAA protein in the culture median (100 ng/mL) or culture median as a control (n=6 per group magnification, 10×). All the data were shown as the mean±SEM of values from six independent experiments. **p<0.01 and ***p<0.001. BLCA, bladder urothelial carcinoma; ChIP, Chromatin immunoprecipitation; DMEM, Dulbecco's Modified Eagle Medium; NNMT, nicotinamide N-methyltransferase; RT-qPCR, quantitative reverse transcription PCR; SAA, serum amyloid A; TCGA, The Cancer Genome Atlas; UBC, urothelial bladder cancer.
Moreover, we found that the addition of methionine to cultured Nnmt-overexpressing NIH/3T3 cells reduced the expression of Saa3 in both ELISA (figure 7C) assay and immunostaining analyses (figure 7D). To confirm that the decrease in H3K27 trimethylation induces the expression of Saa3, we treated NIH/3T3 cells with PF-06726304, which can inhibit H3K27 trimethylation (H3K27me3). Western blotting analysis indicated that the expression of Saa3 increased, while targeting demethylase with the selective pan-KDM5 inhibitor CPI-455 could increase the levels of H3K4 trimethylation (H3K4 me3) in cells and could increase the expression of Saa3 (online supplemental figure S9C). These results suggest that Saa3 expression may be regulated by DNA methylation mechanisms.
Previous studies have indicated that SAA3 is a non-functional pseudogene in the human genome.36 37 Among the three functional SAA family genes (SAA1, SAA2, and SAA4), SAA1 is structurally and functionally similar to the murine Saa3. Therefore, we conducted gene expression correlation analyses between the NNMT gene and the SAA1 gene. In our 32 freshly frozen UBC tissue samples, we found a significant correlation between the NNMT gene and the SAA1 gene (r=0.34, p<0.001; figure 7E). The results were consistent with those found in the TCGA-BLCA dataset (r=0.48, p<0.001; figure 7F). Furthermore, we examined the function of SAA using transwell analysis. We found that SAA significantly enhanced macrophage migration in vitro (68.0±6.77 vs 274.5±6.93, p<0.001; figure 7G). In addition, we confirmed that the expression of SAA1 was positively correlated with that of many pro-tumor macrophage markers (CD163 and CD206) in our 32 UBC tissues and TCGA-BLCA dataset (online supplemental figure S9D,E). These results support the notion that SAA1 may play a role in human bladder cancer, similar to Saa3 in mouse tumors.
NNMT+ CAFs mediated apoptotic signaling pathways and contributed to immunotherapy resistance in patients with UBC
To explore the association between NNMT+ CAF-mediated signaling pathways and immunotherapy resistance in clinical samples, we assayed lymphocyte infiltration and apoptosis of tumor cells in post-treatment tissues from patients with UBC treated with four courses of neoadjuvant tislelizumab combined with cisplatin and gemcitabine. We found that the apoptosis of tumor cells, as markers of Cleaved-Caspase-3 and pan CK double-positive, was increased in tissues with a low proportion of NNMT+ fibroblasts, while the tissues with a high proportion of NNMT+ fibroblasts showed enriched macrophage infiltration (figure 8A–B).
NNMT+ CAFs mediate signaling pathways and contribute to immunotherapy resistance. (A–B) Representative pictures showing the post-treatment stromal tissues staining apoptosis of tumor cells and lymphocyte infiltration from two patients with UBC treated with four courses of neoadjuvant tislelizumab combined with cisplatin and gemcitabine. Magnification, 20×. NNMT+ fibroblasts maker (green), CD8+ T cells marker (CD8, yellow), macrophage marker (CD68, orange), tumor cell marker (CK, blue), apoptosis maker (Cleaved-Caspase 3, red). (C) Α heatmap of 113 excluded phenotype patients demonstrated the top 657 upregulated and 244 downregulated genes following non-responders compared with responders from IMvigor210 dataset (|fold change|>2.0, adjusted p<0.05). (D) GSEA showing that TGF-β receptor signaling, adaptive immune response and leukocyte migration enriched in non-responders (n=85) compared with responders (n=28) from IMvigor210 dataset. NES denoted normalized enrichment score. (E) Violin plot showing the expression NNMT, TGFB1, SAA1, TLR2, IL6, and LIF in non-responders (n=85) compared with responders (n=28) from the IMvigor210 dataset. (F–G) The relative expression of NNMT on human fibroblasts (MRC-5) treated with TGF-β analyzed using RT-qPCR (F) and western blotting analysis (G). The fold change compared with the average value of the control group is presented as the mean±SEM Two-tailed Student’s t-tests were used to evaluate the statistical significance of the differences between two groups. *p<0.05, **p<0.01, and ***p<0.001. CAFs, cancer-associated fibroblasts; CK, cytokeratin; DAPI, 4',6-diamidino-2-phenylindole; GSEA, gene set enrichment analysis; NNMT, nicotinamide N-methyltransferase; TGF, transforming growth factor; UBC, urothelial bladder cancer.
In our bioinformatics analysis of IMvigor210 data, we found that the fibroblast, inflammatory, and TGF-β pathways (figure 8C) were significantly enriched in the non-responders. GSEA analysis also confirmed that immune-related pathways such as TGF-β receptor signaling, adaptive immune response, and leukocyte migration were significantly enriched in non-responders (figure 8D). We found that the genes in the NNMT+ CAF-mediated signaling pathways, such as NNMT, TGFB1, SAA1, IL6, LIF, and TLR2, were significantly elevated in the non-responders compared with the responders (figure 8E). Further experiments showed that TGF-β was a potent regulator of NNMT and SAA1 expression in MRC5 fibroblasts in vitro (figure 8F–G). Thus, these data indicate that the NNMT+ CAFs signaling pathways are associated with resistance to immunotherapy in UBC; combination therapy of NNMT inhibitors with ICI-based immunotherapy may help improve their therapeutic effects in UBC treatment.
Discussion
Non-response to immunotherapy in UBC is associated with substantial genetic and cellular heterogeneity in tumor components and complex interactions among cells in the TME. Here, we report that NNMT+ CAFs were a predictor of non-response in patients with UBC treated with anti-PD-L1 agents and were significantly associated with unfavorable prognosis of UBC. Functionally, elevated NNMT expression in stromal fibroblasts promotes tumor cell progression and immunotherapy resistance via the epigenetic regulatory axis of SAAs in association with CAFs and TAMs.
Previous studies have revealed that some specific subtypes of CAFs are associated with treatment resistance,30 38 39 but few studies have studied the effects of CAFs on immunotherapy resistance in the treatment of UBC. A recent study on pancreatic ductal adenocarcinoma showed that LRRC15+ CAFs were associated with poor response to anti-PD-L1 immunotherapy.15 In this study, we identified NNMT+ CAFs as pivotal factors that influence ICI-based immunotherapy responses in patients with UBC. Targeting NNMT+ CAFs significantly reduced tumor growth and enhanced the antitumor effects of anti-PD-L1 antibody in a bladder cancer mouse model, suggesting that targeting the interactions between CAFs and TAMs might be an efficient approach to ameliorate pro-tumorigenic TME and treat UBC. Therefore, NNMT+ CAFs may serve as superior therapeutic targets for UBC immunotherapy.
The CAF-mediated recruitment of TAMs to the UBC stroma may have a fundamental impact on tumor proliferation and metastasis. It has been established that polarized pro-tumor macrophages secrete various cytokines and chemokines, such as interleukin-6 (IL-6), interleukin-10 (IL-10), TGF-β, vascular endothelial growth factor (VEGF), and C-X-C motif chemokine ligand 12 (CXCL12), to facilitate tumor progression. Interestingly, metabolic reprogramming of TAMs can also contribute to tumor progression. Previous studies have suggested that M2-like TAMs exhibit increased fatty acid consumption and elevated levels of glutamine, which are not only essential for macrophage polarization in a feedback loop but are also particularly beneficial for proliferating tumor cells with a high demand for glutamine.40 41 Consequently, metabolic circuitries, together with various cytokine signaling pathways established between cancer cells and TAMs, may be bidirectional and dictated by the co-evolution of both TAMs and cancer cells.42 43
The interplay of CAF-TAMs is crucial for the development of immunotherapy resistance in UBCs. Previous studies have shown that CAF-mediated metabolism modulates the immunosuppressive functions of TILs, especially macrophages, in the TME.44 45 M2 macrophages can influence antitumor and pro-tumor cells by disrupting natural killer and T cell functions via upregulation of PD-L1 and PD-L2.46 47 In addition, Macrophages stimulate Treg cells expansion.48 Interestingly, in non-small cell lung carcinoma, it has recently been reported that tumor-resident macrophages can enhance the recruitment of Tregs by Ccl17 and Tgfb1, thereby establishing a niche for cancer cell progression.49 This evidence suggests that macrophages play dominant roles in the regulation of immune responses in tumors, and that dysregulation is directly associated with tumor progression and immune evasion. Therefore, elucidating the interplay of CAF-TAMs is crucial for understanding the development of immunotherapy resistance in UBCs.
CAF-TAM crosstalk in the UBC stroma may confer resistance to immunotherapy through a variety of mechanisms mediated by cytokines and chemokines. Previous studies have found that cytokines/chemokines secreted by CAFs broadly affect the recruitment and phenotypic differentiation of immune cells, thereby influencing tumor evolution.44 Specifically, CAF-derived cytokines such as IL-6, IL-8, TGF-β, and macrophage migration inhibitory factor (MIF) have been demonstrated to recruit and induce M2 macrophage differentiation in the TME.44 50 The increased number of macrophages, especially M2 macrophages, in tumor tissues plays a key role in tumor progression.51 In addition, CAFs are the main source of C-C motif chemokine ligand 2 (CCL2) and CXCL12 chemokines, which recruit monocytes to the tumor and facilitate the differentiation of pro-tumor macrophages.52 In this study, we identified a unique subpopulation of fibroblasts (NNMT+ fibroblasts) that was significantly associated with pro-tumor macrophage infiltration in UBC. NNMT+ fibroblasts upregulate SAA expression by decreasing histone H3K27 methylation at the promoter region. The secretion of SAAs in the TME induces the recruitment and polarization of pro-tumor macrophages to facilitate tumor progression. The mechanisms of CAF-derived SAA signaling may explain the recruitment of macrophages and their pro-tumorigenic effects in our clinical samples.
Saa3 belongs to the SAA gene family. Previous studies have suggested that Saa3 is expressed in macrophages and fibroblasts, and is effectively induced by Il-1β, Tnfα, and Il-6 via NF-κB signaling in mice.36 53 In this study, we identified a novel regulatory mechanism of Saa3 involving decreasing histone methylation by upregulating NNMT in fibroblasts. A previous study showed that conditional KO of Saa3 in CAFs reduced fibrosis and infiltration of macrophages into the stroma.53 It had been shown that Saa3 was required for the pro-tumorigenic properties of CAFs. In the present study, we found that Saa3 activated macrophage NF-κB signaling to shift them towards an M2-like phenotype and enhanced Saa3 expression via a positive feedback loop. Notably, Saa3 exerts its effects on macrophages via TLR2/TLR4 and polarizes TAMs into tumor-promoting phenotypes.37 54 In addition to macrophages, CAF-secreted SAAs may regulate the recruitment and differentiation of other immune cells, such as myeloid-derived suppressive cells.55 The regulation of SAA signaling by NNMT+CAFs to modulate immunotherapy in UBC should be explored in future studies.
The CSF1-CSF1R axis in fibroblast-macrophage interactions has been well-documented in previous studies.56 57 Our findings on NNMT+CAFs-mediated SAA1 signaling networks that regulate macrophages may add more mechanistic modules between CAFs and TAMs. However, the mechanisms by which SAA proteins recruit monocytes have yet to be fully elucidated in our study. A previous study58 demonstrated that SAA proteins trigger a chemotactic cascade to prolong leukocyte recruitment to inflammation sites. Specifically, SAA1 attracts monocyte-derived immature dendritic cells (DCs) by rapidly inducing chemokine in monocytes and DCs. Interestingly, the study uncovered that SAA chemoattracts immune cells via a G-protein-coupled receptor. Nevertheless, further investigation is needed to characterize the mechanisms by which SAA proteins influence immune cell migration in UBC.
Extracellular matrix (ECM) remodeling cascade triggered by cytokines and chemokines might also confer significant impacts on resistance to immunotherapy of UBC. Various factors, including TMB, pre-existing T-cell immunity, and the signature of TGF-β, have been reported to impact the outcome of immunotherapy of UBC.14 We observed that NNMT significantly induced morphological changes in fibroblasts in vitro, which is consistent with the results of a previous study showing that NNMT promotes collagen contractility in fibroblasts.22 Notably, CAFs can synthesize and remodel the extracellular matrix as a physical barrier against immune infiltration, rendering them resistant to immunotherapy, particularly in immune-excluded phenotypes;13 and TAMs could play a pivotal role in various tumor collagens to remodel the extracellular matrix.59 Altogether, CAFs-mediated and TAMs-mediated ECM remodeling may contribute to the development of the exclusive immune subtype of UBC and eventually confer resistance to immunotherapy.
Nevertheless, the potential molecular mechanisms of TAMs might be involved in regulating CD8+ T cell-mediated cancer cell killing. CAFs and TAMs could directly render T-cell dysfunction and immunotherapy resistance, for example, by upregulating the expression of immune checkpoint V-domain Ig suppressor of T cell activation (VISTA), which is activated by histamine receptor H1 (HRH1).26 Notably, one of our concurrent studies revealed that depleting macrophages in MB49 wild type (WT) tumors significantly reduces T-cell infiltration, suggesting a key role for TAMs in inducing immunosuppression within the TME,60 enhancing our understanding of the roles of TAMs in the regulation of T cells in UBC. Altogether, the regulatory axis of NNMT+ CAFs might facilitate cell-cell interactions between TAMs and T cells, involving the interactions of immunosuppressive molecules. However, further research is necessary to fully elucidate these mechanisms.
Understanding the regulatory mechanisms underlying CAF-TAMs is important in immunotherapy. It has been reported that targeting TAMs signaling to increase the M1-like polarized macrophages using a TLR7/8 agonist was a promising approach to develop novel therapeutics for cancer treatment.27 Notably, we found that inhibiting NNMT+ fibroblasts in UBC models reduced macrophage infiltration and enhanced the antitumor effects of the anti-PD-L1 antibody. Therefore, targeting the CAF-TAM crosstalk might be an efficient approach to improve the immunotherapeutic response by ameliorating the immunosuppressive TME and/or promoting T cell infiltration into intra-tumor areas.
As for the limitations of our study, although we observed that NNMTi was able to enhance the reduction of UBC tumor growth when treated with anti-PD-L1 agents in mouse models, using a design with large samples of an organoid model may allow us to clarify the synergistic effect of the combination of NNMTi and anti-PD-L1 agents in UBC immunotherapy. Moreover, the sample size used to analyze the ICI treatment responses was relatively small. In future translational research, clinical analyses in large cohorts will be crucial to validate the association between NNMT+ CAFs and ICI treatment responses. Finally, for the molecular mechanisms of NNMT+ CAFs, comprehensive analyses of the lymphocytic compartment in tumors might provide further understanding of the interactions between NNMT+ CAFs with other immune cells. Integrated scRNA sequencing with spatial transcriptomic analyses of cell populations in UBC tissues in situ may help elucidate the interactions between NNMT+ CAFs and immune cells for the development of treatment resistance to immunotherapy.
Conclusions
The present study revealed that NNMT+ CAFs are associated with non-response to PD-L1 blockade immunotherapy in patients with UBC. Elevated NNMT specifically in CAFs upregulates SAA expression and secretion in the tumor microenvironment, causes increased recruitment of macrophages, and induces macrophage differentiation and polarization, thereby directly or indirectly promoting tumor progression and rendering UBC resistant to immunotherapy. Targeting NNMT+ CAFs may be a promising novel therapeutic approach for tackling immunotherapy resistance in bladder cancer.
Data availability statement
Data are available in a public, open access repository. Data sharing not applicable as no datasets generated and/or analyzed for this study.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and all human samples were anonymously coded following the local ethical guidelines as stipulated by the Declaration of Helsinki. Written informed consent was obtained from all patients, and the use of clinical information and human tissues from the cohorts in this study was approved by the Clinical Research Ethics Committee of Sun Yat-sen Memorial Hospital, Sun Yat-sen University (Approval Number SYSKY -2022–011-01). Participants gave informed consent to participate in the study before taking part.
Acknowledgments
We are grateful to all patients and their families and the staff at Sun Yat-sen University, who helped us to accomplish this study.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
X @yangmeihua1
MY, BW and WH contributed equally.
Contributors MY, BW, and WH designed the experiment, evaluated the immunohistochemical markers of tissue sections, analyzed the data, and wrote the manuscript. HZ analyzed the single-cell RNA data. WHe and X-KZ collected the samples and clinical data. DY and HY performed immunohistochemical staining. LH, LP, and KL contributed to revising the manuscript. HQ contributed to data interpretation, writing, and revision of the manuscript. TL and JH supervised the study design and finalized the manuscript. JH as the guarantor for the overall content, approved and supervised the project. All authors have read and approved the manuscript.
Funding This study was supported by the National Key Research and Development Program of China (Grant No. 2018YFA0902803), the National Natural Science Foundation of China (Grant No. 81825016, 81772719, 8217326, 8217366, 81972731, 81773026, 81772728, 81972385, 81961128027 and 82002679), The Key Areas Research and Development Program of Guangdong (Grant No. 2018B010109006), the Science and Technology Planning Project of Guangdong Province (Grant No. 2017B020227007), Project Supported by Guangdong Province Higher Vocational Colleges & Schools Pearl River Scholar Funded Scheme (for Tianxin Lin), Guangdong Provincial Clinical Research Center for Urological Diseases (2020B1111170006), the Key Laboratory of Malignant Tumor Molecular Mechanism and Translational Medicine of Guangzhou Bureau of Science and Information Technology (Grant No. 013-163), and the Young Talents Fund of The First Affiliated Hospital of Nanchang University (YFYPY202260).
Competing interests No, there are no competing interests.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.